4. Description-based simulation

In lesson 3 the events were generated by the layout. They can also come from an existing sequence: a Pulseq file written with another tool, or the sequence description a scanner streams during acquisition. This lesson simulates such descriptions directly, without restating the sequence parameters.

Learning objectives

• Inspect a sequence description: events, timestamps and RF definitions.

• Simulate a description with from_description, and understand the role of the handlers.

• Simulate a Pulseq file with from_pulseq.

Previous: 3. Custom signal model. Next: 5. Custom operator, which adds a custom preparation and readout.

Sequence description

A sequence description (SequenceDescription) lists the events of one repetition: RF pulses, ADC windows and waits, each with its timestamp and parameters (flip angle, phase, RF definition). It is the input of every simulation.

The default description of a simulator is generated by its layout from the sequence parameters, and describe returns it. For the 60-degree FSE of lesson 1:

import torch

from blochsim.simulators import FSESimulator

fse = FSESimulator(ESP=5.0, TR=3000.0)
flip = torch.full((48,), 60.0)

description = fse.describe(flip=flip, ESP=5.0, TR=3000.0)
print(f"{len(description.events)} events, {len(description.adc_events)} readouts")

145 events, 48 readouts

The first 30 ms show the excitation and the refocusing pulses centred between readouts. The description contains no gradients: the scanner stream does not carry them, and blochsim does not read them from a Pulseq file.

Simulation from a description

from_description constructs a simulator from a description, bypassing the layout. Only tissue properties and simulation settings are passed:

from_events = FSESimulator.from_description(description, states=48, T1=1000.0, T2=80.0)
signal = from_events.simulate()

The handlers of the class are still used: they play each event and define the dephasing between events, which the description does not specify. FSESimulator places crushers around each refocusing pulse. The same events played with MRFSimulator, whose readout applies one unit of dephasing after each sample, describe a different sequence:

from blochsim.simulators import MRFSimulator

as_mrf = MRFSimulator.from_description(description, states=48, T1=1000.0, T2=80.0)
wrong = as_mrf.simulate()

With the FSE handlers the result matches the shipped simulator up to a small offset: FSESimulator also applies the T1 recovery between trains, which is not an event and is therefore absent from the description. With the MRF handlers the CPMG condition is not met and the echo amplitudes oscillate and decay. The class therefore specifies how the events are to be interpreted.

Scanner stream

During acquisition, pulserver streams the description of the running sequence with the raw data. read_mrd_description() reads it from an MRD stream, and from_description simulates it as above. A reconstruction can thus simulate the sequence that was actually played.

Pulseq file

from_pulseq reads the description from a .seq file, parsed by pypulseq. The file below is a short FSE:

from_file = FSESimulator.from_pulseq(SEQ_FILE, states=20)
file_signal = from_file.simulate(T1=1000.0, T2=torch.tensor([80.0, 110.0]))

8 echoes, 6.3 ms apart

Echo spacing, echo train length, refocusing angle and RF shapes are read from the file; only the tissue and the handler class are specified.

The number of blocks per repetition is taken from the TRSize definition of the file. Each readout is timestamped at the sample where the trajectory crosses the k-space centre, i.e. at the echo time.

Specification

What this lesson did, stated the way you would ask an agent for it:

With blochsim, read the sequence in fse.seq with FSESimulator.from_pulseq
and simulate it in white matter and grey matter (T1 1 s; T2 80 and
110 ms). Plot the events of one repetition and the signal at each echo
time. Separately, take the description of a 48-echo, 60-degree FSE and
simulate it with the FSE handlers and with the MRF handlers, to show that
the handlers decide what happens between the events.